MSE period based estimation of first quantization step in double compressed JPEG images

- A mean square error (MSE) sequence based approach is proposed for the first quantization step q1 estimation in double compressed JPEG images.
- A Histogram adjustment strategy is proposed to enhance the periodic effect.
- A strategy of histogram comparison is proposed to estimate q1 from the candidates.
- Experimental results demonstrate that the approach outperforms the state of the art.

The estimation of the first quantization step in double JPEG compressed images is still a challenging problem, especially when the first quantization step q1 is smaller than the second quantization step q2. In this paper, we present a novel method to estimate q1. By introducing the mean square error (MSE) sequence of ratios among DCT coefficient histogram bins, we formulate the relationship between its periodic fluctuation and q1. And in order to enhance the periodic effect, we propose a strategy to adjust the histogram. Then, based on MSE sequence, several q1 candidates can be obtained. Finally by histogram comparison, the estimated quantization step is selected from the candidates. Experimental results demonstrate that the proposed approach has better overall performance when compared with state of the art methods.